[2/N][torch.compile] make compilation cfg part of vllm cfg (#10383)

Signed-off-by: youkaichao <youkaichao@gmail.com>

tests/compile/piecewise/test_simple.py

@@ -11,8 +11,8 @@ from torch.library import Library
 from vllm.compilation.compile_context import set_compile_context
 from vllm.compilation.counter import compilation_counter
 from vllm.compilation.decorators import support_torch_compile
-from vllm.compilation.levels import CompilationLevel
-from vllm.config import VllmConfig
+from vllm.config import CompilationLevel, VllmConfig
+from vllm.plugins import set_current_vllm_config
 from vllm.utils import direct_register_custom_op
 
 global_counter = 0
@@ -82,7 +82,9 @@ def test_simple_piecewise_compile():
     os.environ["VLLM_TORCH_COMPILE_CONFIG"] = config
     os.environ["VLLM_TORCH_COMPILE_LEVEL"] = str(CompilationLevel.PIECEWISE)
 
-    model = SillyModel(vllm_config=VllmConfig(), prefix='')
+    vllm_config = VllmConfig()
+    with set_current_vllm_config(vllm_config):
+        model = SillyModel(vllm_config=vllm_config, prefix='')
 
     inputs = torch.randn(100).cuda()
 

tests/compile/piecewise/test_toy_llama.py

@@ -15,12 +15,10 @@ from torch import nn
 from torch.library import Library
 
 from vllm.compilation.compile_context import set_compile_context
-from vllm.compilation.config import CompilationConfig
 from vllm.compilation.counter import compilation_counter
 from vllm.compilation.decorators import support_torch_compile
-from vllm.compilation.levels import CompilationLevel
-from vllm.config import VllmConfig
-from vllm.plugins import set_compilation_config
+from vllm.config import CompilationConfig, CompilationLevel, VllmConfig
+from vllm.plugins import set_compilation_config, set_current_vllm_config
 from vllm.utils import direct_register_custom_op
 
 # create a library to hold the custom op
@@ -272,9 +270,11 @@ def run_model(llama_config,
             CompilationLevel.NO_COMPILATION)
         set_compilation_config(None)
 
-    model = LlamaModel(config=llama_config,
-                       vllm_config=VllmConfig(),
-                       prefix="").eval().cuda()
+    vllm_config = VllmConfig()
+    with set_current_vllm_config(vllm_config):
+        model = LlamaModel(config=llama_config,
+                           vllm_config=vllm_config,
+                           prefix="").eval().cuda()
 
     B = 16  # max batch size
     input_ids = torch.randint(0, llama_config.vocab_size, (B, )).cuda()
@@ -395,9 +395,11 @@ def benchmark():
         else:
             set_compilation_config(None)
 
-        model = LlamaModel(config=llama_config,
-                           vllm_config=VllmConfig(),
-                           prefix="").eval().cuda().to(torch.bfloat16)
+        vllm_config = VllmConfig()
+        with set_current_vllm_config(vllm_config):
+            model = LlamaModel(config=llama_config,
+                               vllm_config=vllm_config,
+                               prefix="").eval().cuda().to(torch.bfloat16)
 
         B = 256  # max batch size
         input_ids = torch.randint(0, llama_config.vocab_size, (B, )).cuda()

tests/compile/test_basic_correctness.py

@@ -3,7 +3,7 @@ from typing import Dict, List, Optional
 
 import pytest
 
-from vllm.compilation.levels import CompilationLevel
+from vllm.config import CompilationLevel
 from vllm.utils import cuda_device_count_stateless
 
 from ..utils import compare_all_settings

tests/compile/test_full_graph.py

 import pytest
 
-from vllm.compilation.levels import CompilationLevel
+from vllm.config import CompilationLevel
 
 from ..utils import fork_new_process_for_each_test
 from .utils import TEST_MODELS, check_full_graph_support

tests/compile/test_fusion.py

@@ -3,10 +3,10 @@ import torch
 from compressed_tensors.quantization import FP8_DTYPE
 
 import vllm.envs as envs
-from vllm.compilation.config import CompilationConfig
 from vllm.compilation.fusion import (FusionPass, find_auto_fn,
                                      find_auto_fn_maybe)
 from vllm.compilation.reshapes import RedundantReshapesPass
+from vllm.config import CompilationConfig
 from vllm.model_executor.layers.layernorm import RMSNorm
 from vllm.model_executor.layers.quantization.utils.w8a8_utils import (
     apply_fp8_linear)

tests/compile/test_wrapper.py

@@ -3,6 +3,7 @@ from typing import Optional
 import torch
 
 from vllm.compilation.wrapper import TorchCompileWrapperWithCustomDispatcher
+from vllm.config import CompilationLevel
 
 
 class MyMod(torch.nn.Module):
@@ -18,7 +19,8 @@ class MyWrapper(TorchCompileWrapperWithCustomDispatcher):
     def __init__(self, model):
         self.model = model
         compiled_callable = torch.compile(self.forward, backend="eager")
-        super().__init__(compiled_callable)
+        super().__init__(compiled_callable,
+                         compilation_level=CompilationLevel.DYNAMO_ONCE)
 
     def forward(self, x: torch.Tensor, cache: Optional[torch.Tensor] = None):
         # this is the function to be compiled

tests/compile/utils.py

@@ -4,7 +4,7 @@ import torch
 
 from tests.quantization.utils import is_quant_method_supported
 from vllm import LLM, SamplingParams
-from vllm.compilation.levels import CompilationLevel
+from vllm.config import CompilationLevel
 from vllm.platforms import current_platform
 
 TEST_MODELS = [

tests/model_executor/test_enabled_custom_ops.py

@@ -3,11 +3,13 @@ from typing import List
 
 import pytest
 
+from vllm.config import CompilationConfig, VllmConfig
 from vllm.model_executor.custom_op import CustomOp
 from vllm.model_executor.layers.activation import (GeluAndMul,
                                                    ReLUSquaredActivation,
                                                    SiluAndMul)
 from vllm.model_executor.layers.layernorm import RMSNorm
+from vllm.plugins import set_current_vllm_config
 
 
 # Registered subclass for test
@@ -51,42 +53,40 @@ class Relu3(ReLUSquaredActivation):
     ])
 def test_enabled_ops(env: str, torch_level: int, ops_enabled: List[int],
                      default_on: bool):
-    os.environ["VLLM_CUSTOM_OPS"] = env
     os.environ["VLLM_TORCH_COMPILE_LEVEL"] = str(torch_level)
+    vllm_config = VllmConfig(compilation_config=CompilationConfig(
+        custom_ops=env.split(",")))
+    with set_current_vllm_config(vllm_config):
+        assert CustomOp.default_on() == default_on
 
-    # Reset default_on (computed once):
-    CustomOp.default_on.cache_clear()
+        ops_enabled = [bool(x) for x in ops_enabled]
 
-    assert CustomOp.default_on() == default_on
+        assert RMSNorm(1024).enabled() == ops_enabled[0]
+        assert CustomOp.op_registry["rms_norm"].enabled() == ops_enabled[0]
 
-    ops_enabled = [bool(x) for x in ops_enabled]
+        assert SiluAndMul().enabled() == ops_enabled[1]
+        assert CustomOp.op_registry["silu_and_mul"].enabled() == ops_enabled[1]
 
-    assert RMSNorm(1024).enabled() == ops_enabled[0]
-    assert CustomOp.op_registry["rms_norm"].enabled() == ops_enabled[0]
+        assert GeluAndMul().enabled() == ops_enabled[2]
+        assert CustomOp.op_registry["gelu_and_mul"].enabled() == ops_enabled[2]
 
-    assert SiluAndMul().enabled() == ops_enabled[1]
-    assert CustomOp.op_registry["silu_and_mul"].enabled() == ops_enabled[1]
+        # If registered, subclasses should follow their own name
+        assert Relu3().enabled() == ops_enabled[3]
+        assert CustomOp.op_registry["relu3"].enabled() == ops_enabled[3]
 
-    assert GeluAndMul().enabled() == ops_enabled[2]
-    assert CustomOp.op_registry["gelu_and_mul"].enabled() == ops_enabled[2]
+        # Unregistered subclass
+        class SiluAndMul2(SiluAndMul):
+            pass
 
-    # If registered, subclasses should follow their own name
-    assert Relu3().enabled() == ops_enabled[3]
-    assert CustomOp.op_registry["relu3"].enabled() == ops_enabled[3]
-
-    # Unregistered subclass
-    class SiluAndMul2(SiluAndMul):
-        pass
-
-    # Subclasses should not require registration
-    assert SiluAndMul2().enabled() == SiluAndMul().enabled()
+        # Subclasses should not require registration
+        assert SiluAndMul2().enabled() == SiluAndMul().enabled()
 
 
 @pytest.mark.parametrize(
     "env", ["all,none", "all,+rms_norm,all", "+rms_norm,-rms_norm"])
 def test_enabled_ops_invalid(env: str):
-    os.environ["VLLM_CUSTOM_OPS"] = env
-    CustomOp.default_on.cache_clear()
-
-    with pytest.raises(AssertionError):
-        RMSNorm(1024).enabled()
+    with pytest.raises(Exception):  # noqa
+        vllm_config = VllmConfig(compilation_config=CompilationConfig(
+            custom_ops=env.split(",")))
+        with set_current_vllm_config(vllm_config):
+            RMSNorm(1024).enabled()

tests/tpu/test_compilation.py

@@ -5,7 +5,7 @@ import tempfile
 
 import depyf
 
-from vllm.compilation.levels import CompilationLevel
+from vllm.config import CompilationLevel
 
 # disable custom dispatcher, let Dynamo takes over
 # all the control

tests/tpu/test_custom_dispatcher.py

 import os
 
-from vllm.compilation.levels import CompilationLevel
+from vllm.config import CompilationLevel
 
 from ..utils import compare_two_settings
 

vllm/compilation/backends.py

@@ -10,13 +10,12 @@ import torch
 import torch.fx as fx
 
 import vllm.envs as envs
+from vllm.config import CompilationConfig, CompilationLevel
 from vllm.logger import init_logger
 from vllm.utils import combine_fx_passes, weak_ref_tensors
 
-from .config import CompilationConfig
 from .counter import compilation_counter
 from .fusion import FusionPass
-from .levels import CompilationLevel
 from .reshapes import RedundantReshapesPass
 
 logger = init_logger(__name__)
@@ -392,7 +391,10 @@ class VllmBackend:
     sym_tensor_indices: List[int]
     input_buffers: List[torch.Tensor]
 
-    def __init__(self, post_grad_passes: Sequence[Callable] = ()):
+    def __init__(
+        self,
+        compilation_configs: CompilationConfig,
+    ):
         global global_graph_pool
         if global_graph_pool is None:
             global_graph_pool = torch.cuda.graph_pool_handle()
@@ -401,11 +403,13 @@ class VllmBackend:
         # streams, it might not be safe to share a global pool.
         # only investigate this when we use multiple streams
         self.graph_pool = global_graph_pool
-        self.post_grad_passes = post_grad_passes
+        self.post_grad_passes = []
 
         self.sym_tensor_indices = []
         self.input_buffers = []
 
+        self.compilation_configs = compilation_configs
+
         # `torch.compile` is JIT compiled, so we don't need to
         # do anything here
 
@@ -437,10 +441,10 @@ class VllmBackend:
         assert not self._called, "VllmBackend can only be called once"
 
         self.graph = graph
-        # config is read now, because only here can
+        # config is updated now, because only here can
         # we get the sizes to capture for cudagraph
         # from compilation context
-        self.compilation_configs = CompilationConfig.select_and_init_config()
+        self.compilation_configs.init_during_runtime()
         self.add_passes_to_config()
 
         self.split_gm, self.piecewise_graphs = split_graph(
@@ -688,4 +692,6 @@ def select_default_backend(level: int) -> Union[str, Callable]:
         return backend_str
     assert level == CompilationLevel.PIECEWISE
 
-    return VllmBackend()
+    from vllm.plugins import get_current_vllm_config
+    compilation_config = get_current_vllm_config().compilation_config
+    return VllmBackend(compilation_config)

vllm/compilation/config.py

-import copy
-from pathlib import Path
-from typing import Any, Dict, List, Optional
-
-from pydantic import BaseModel, Field, PrivateAttr
-
-import vllm.envs as envs
-from vllm.logger import init_logger
-
-from .compile_context import get_compile_context
-
-logger = init_logger(__name__)
-
-
-class CompilationConfig(BaseModel):
-    """
-    Configuration for compilation.
-    It has two parts:
-    - CudaGraph capture:
-        - use_cudagraph: whether to use cudagraph inside compilation.
-            - False: cudagraph inside compilation is not used.
-            - True: cudagraph inside compilation is used. It requires
-                that all input buffers have fixed addresses.
-            Note that this is orthogonal to the cudagraph capture out
-            side of compilation.
-            TODO: move outside cudagraph logic into compilation.
-            torch.compile will handle cudagraph capture logic in the future.
-        - cudagraph_capture_sizes: sizes to capture cudagraph.
-            - None: capture sizes are inferred from compilation context.
-            - List[int]: capture sizes are specified.
-        - cudagraph_num_of_warmups: number of warmup runs for cudagraph.
-            It means the first several runs will be treated as warmup runs.
-            Only after that, the execution will be recorded, and the recorded
-            cudagraph will be used for subsequent runs.
-        - cudagraph_copy_inputs: whether to copy input tensors for
-            cudagraph. If the caller can guarantee that the same input buffers
-            are always used, it can set this to False. Otherwise, it should
-            set this to True, and the compiler will copy the input to an
-            internally managed buffer. Default is False.
-    - Inductor compilation:
-        - use_inductor: whether to use inductor compilation.
-            - False: inductor compilation is not used. graph runs in eager.
-            - True: inductor compilation is used. one graph for symbolic shape
-                is compiled. In addition, compile for different sizes specified
-                in inductor_compile_sizes, using configurations
-                in inductor_compile_config.
-        - inductor_compile_sizes: sizes to compile for inductor.
-        - inductor_specialize_for_cudagraph_no_more_than: an optional integer
-            to specialize inductor for cudagraph sizes no more than the
-            specified size. It is useful when we want to specialize inductor
-            with a subset of cudagraph sizes.
-        - inductor_compile_config: additional configurations for inductor.
-            - None: use default configurations.
-        - inductor_passes: additional passes for inductor. It is a dictionary
-            from pass name to pass function qualified name. We use function
-            name because the config uses json format. If we pass the config
-            from Python, functions can also be passed directly via Python object
-            constructor, e.g. `CompilationConfig(inductor_passes={"a": func})`
-    - Custom inductor passes:
-        - dump_graph_stages: list of stages for which we want to dump the graph.
-            Each pass defines its own stages (before, after, maybe in-between).
-        - dump_graph_dir: directory to dump the graph. Default is .
-        - enable_fusion: whether to enable the custom fusion pass.
-            TODO better pass enabling system.
-    
-    Why we have different sizes for cudagraph and inductor:
-    - cudagraph: a cudagraph captured for a specific size can only be used
-        for the same size. We need to capture all the sizes we want to use.
-    - inductor: a graph compiled by inductor for a general shape can be used
-        for different sizes. Inductor can also compile for specific sizes,
-        where it can have more information to optimize the graph with fully
-        static shapes. However, we find the general shape compilation is
-        sufficient for most cases. It might be beneficial to compile for
-        certain small batchsizes, where inductor is good at optimizing.
-    """
-    use_inductor: bool = True
-    inductor_specialize_for_cudagraph_no_more_than: Optional[int] = None
-    inductor_compile_sizes: Optional[List[int]] = Field(default_factory=dict)
-    inductor_compile_config: Dict = Field(default_factory=dict)
-    inductor_passes: Dict[str, str] = Field(default_factory=dict)
-
-    use_cudagraph: bool = False
-    non_cudagraph_ops: List[str] = Field(default_factory=list)
-    cudagraph_num_of_warmups: int = 0
-    cudagraph_capture_sizes: Optional[List[int]] = None
-    cudagraph_copy_inputs: bool = False
-
-    dump_graph_stages: List[str] = Field(default_factory=list)
-    dump_graph_dir: Path = Field(default=Path("."))
-    enable_fusion: bool = True
-
-    # not configurable, computed after init
-    compile_sizes: List[int] = PrivateAttr
-    capture_sizes: List[int] = PrivateAttr
-
-    def model_post_init(self, __context: Any) -> None:
-        for k, v in self.inductor_passes.items():
-            if not isinstance(v, str):
-                assert callable(v), (
-                    f"pass {k} should be a function or a qualified name")
-                self.inductor_compile_config[k] = v
-                continue
-
-            # resolve function from qualified name
-            names = v.split(".")
-            module = ".".join(names[:-1])
-            func_name = names[-1]
-            func = __import__(module).__dict__[func_name]
-            self.inductor_compile_config[k] = func
-
-    def init_during_runtime(self):
-        """To complete the initialization of config,
-        we need to know the compile context, which is only available
-        during the first run of the model.
-        """
-        context = get_compile_context()
-        context = copy.deepcopy(context) if context is not None else []
-        sizes_to_specialize: List[int] = context
-        if self.cudagraph_capture_sizes is None:
-            self.capture_sizes = sizes_to_specialize
-        else:
-            self.capture_sizes = self.cudagraph_capture_sizes
-            logger.info(("cudagraph sizes specified by model runner"
-                         " %s is overridden by config %s"),
-                        sizes_to_specialize, self.cudagraph_capture_sizes)
-        if self.inductor_specialize_for_cudagraph_no_more_than is not None:
-            assert self.inductor_compile_sizes is None, (
-                "inductor_compile_sizes should be None when "
-                "inductor_specialize_for_cudagraph_no_more_than is not None")
-            self.compile_sizes = [
-                x for x in self.capture_sizes
-                if x <= self.inductor_specialize_for_cudagraph_no_more_than
-            ]
-        else:
-            assert self.inductor_compile_sizes is not None, (
-                "inductor_compile_sizes should not be None when "
-                "inductor_specialize_for_cudagraph_no_more_than is None")
-            self.compile_sizes = self.inductor_compile_sizes
-
-    @staticmethod
-    def select_and_init_config() -> "CompilationConfig":
-        """The order of selecting config is:
-        1. Use the config specified in environment variable.
-        2. Use the config specified in plugins.
-        3. Use the default config.
-        """
-        config_path = envs.VLLM_TORCH_COMPILE_CONFIG
-        if config_path is not None:
-            with open(config_path) as json_file:
-                config = CompilationConfig.model_validate_json(
-                    json_file.read())
-        else:
-            from vllm.plugins import get_compilation_config
-            predefined_config = get_compilation_config()
-            config = predefined_config if predefined_config is not None else (
-                CompilationConfig())
-
-        config.init_during_runtime()
-        return config

vllm/compilation/decorators.py

@@ -3,10 +3,8 @@ from typing import Dict, List, Optional, Union
 
 import torch
 
-import vllm.envs as envs
-from vllm.compilation.levels import CompilationLevel
 from vllm.compilation.wrapper import TorchCompileWrapperWithCustomDispatcher
-from vllm.config import VllmConfig
+from vllm.config import CompilationLevel, VllmConfig
 from vllm.logger import init_logger
 from vllm.sequence import IntermediateTensors
 from vllm.utils import supports_dynamo
@@ -126,12 +124,14 @@ def _support_torch_compile(cls: type,
         old_init(self, vllm_config=vllm_config, prefix=prefix, **kwargs)
         # for CompilationLevel.DYNAMO_AS_IS , the upper level model runner
         # will handle the compilation, so we don't need to do anything here.
-        self.do_not_compile = envs.VLLM_TORCH_COMPILE_LEVEL in [
+        self.do_not_compile = \
+            vllm_config.compilation_config.level in [
             CompilationLevel.NO_COMPILATION, CompilationLevel.DYNAMO_AS_IS
         ] or not supports_dynamo()
         if self.do_not_compile:
             return
-        TorchCompileWrapperWithCustomDispatcher.__init__(self)
+        TorchCompileWrapperWithCustomDispatcher.__init__(
+            self, compilation_level=vllm_config.compilation_config.level)
 
     cls.__init__ = __init__  # type: ignore
 

vllm/compilation/fusion.py

@@ -6,8 +6,8 @@ from torch._higher_order_ops.auto_functionalize import auto_functionalized
 from torch._inductor.pattern_matcher import (Match, PatternMatcherPass,
                                              fwd_only, register_replacement)
 
-from vllm.compilation.config import CompilationConfig
 from vllm.compilation.inductor_pass import InductorPass
+from vllm.config import CompilationConfig
 from vllm.logger import init_logger
 
 logger = init_logger(__name__)

vllm/compilation/inductor_pass.py

@@ -2,7 +2,7 @@ from abc import ABC, abstractmethod
 
 import torch
 
-from vllm.compilation.config import CompilationConfig
+from vllm.config import CompilationConfig
 # yapf: disable
 from vllm.distributed import get_tensor_model_parallel_rank as get_tp_rank
 from vllm.distributed import (

vllm/compilation/levels.py

-# constants for the levels of the compilation process
-
-
-class CompilationLevel:
-    NO_COMPILATION = 0
-    DYNAMO_AS_IS = 1
-    DYNAMO_ONCE = 2
-    PIECEWISE = 3

vllm/compilation/wrapper.py

@@ -8,8 +8,7 @@ from typing import Callable, List, Optional
 import torch
 
 import vllm.envs as envs
-
-from .levels import CompilationLevel
+from vllm.config import CompilationLevel
 
 
 class TorchCompileWrapperWithCustomDispatcher:
@@ -25,7 +24,9 @@ class TorchCompileWrapperWithCustomDispatcher:
         `torch.compile` over the forward method.
     """
 
-    def __init__(self, compiled_callable: Optional[Callable] = None):
+    def __init__(self,
+                 compiled_callable: Optional[Callable] = None,
+                 compilation_level: int = 0):
 
         if compiled_callable is None:
             # default compilation settings
@@ -38,7 +39,7 @@ class TorchCompileWrapperWithCustomDispatcher:
             backend = get_torch_compile_backend()
             if backend is None:
                 from vllm.compilation.backends import select_default_backend
-                backend = select_default_backend(envs.VLLM_TORCH_COMPILE_LEVEL)
+                backend = select_default_backend(compilation_level)
 
             compiled_callable = torch.compile(
                 self.forward,
@@ -54,7 +55,7 @@ class TorchCompileWrapperWithCustomDispatcher:
         # subclasses can use this to switch between the custom dispatcher
         # and the default Dynamo guard mechanism.
         self.use_custom_dispatcher: bool = \
-            envs.VLLM_TORCH_COMPILE_LEVEL >= CompilationLevel.DYNAMO_ONCE
+            compilation_level >= CompilationLevel.DYNAMO_ONCE
 
     def __call__(self, *args, **kwargs):
         """Implement the dispatch logic here, beyond the torch.compile level.

vllm/config.py

@@ -3,10 +3,12 @@ import enum
 import json
 import warnings
 from dataclasses import dataclass, field, replace
+from pathlib import Path
 from typing import (TYPE_CHECKING, Any, Callable, ClassVar, Dict, Final, List,
                     Literal, Mapping, Optional, Set, Tuple, Type, Union)
 
 import torch
+from pydantic import BaseModel, Field, PrivateAttr
 from transformers import PretrainedConfig
 
 import vllm.envs as envs
@@ -2052,6 +2054,185 @@ class ObservabilityConfig:
                 f"installed. Original error:\n{otel_import_error_traceback}")
 
 
+class CompilationLevel:
+    # constants for the levels of the compilation process
+    NO_COMPILATION = 0
+    DYNAMO_AS_IS = 1
+    DYNAMO_ONCE = 2
+    PIECEWISE = 3
+
+
+class CompilationConfig(BaseModel):
+    """
+    Configuration for compilation.
+    It has three parts:
+    - Top-level Compilation control:
+        - level: the level of compilation.
+            - 0: no compilation.
+            - 1: dynamo as is.
+            - 2: dynamo once.
+            - 3: piecewise compilation.
+        - custom_ops: fine-grained control over which custom ops to enable/disable.
+            Use 'all' to enable all, 'none' to disable all.
+            Also specify a list of custom op names to enable (prefixed with a '+'),
+            or disable (prefixed with a '-').
+            Examples:
+                - 'all,-op1' to enable all except op1
+                - 'none,+op1,+op2' to enable only op1 and op2
+            By default, all custom ops are enabled when running without Inductor
+                and disabled when running with Inductor (compile_level >= Inductor).
+    - CudaGraph capture:
+        - use_cudagraph: whether to use cudagraph inside compilation.
+            - False: cudagraph inside compilation is not used.
+            - True: cudagraph inside compilation is used. It requires
+                that all input buffers have fixed addresses.
+            Note that this is orthogonal to the cudagraph capture out
+            side of compilation.
+            TODO: move outside cudagraph logic into compilation.
+            torch.compile will handle cudagraph capture logic in the future.
+        - cudagraph_capture_sizes: sizes to capture cudagraph.
+            - None: capture sizes are inferred from compilation context.
+            - List[int]: capture sizes are specified.
+        - cudagraph_num_of_warmups: number of warmup runs for cudagraph.
+            It means the first several runs will be treated as warmup runs.
+            Only after that, the execution will be recorded, and the recorded
+            cudagraph will be used for subsequent runs.
+        - cudagraph_copy_inputs: whether to copy input tensors for
+            cudagraph. If the caller can guarantee that the same input buffers
+            are always used, it can set this to False. Otherwise, it should
+            set this to True, and the compiler will copy the input to an
+            internally managed buffer. Default is False.
+    - Inductor compilation:
+        - use_inductor: whether to use inductor compilation.
+            - False: inductor compilation is not used. graph runs in eager.
+            - True: inductor compilation is used. one graph for symbolic shape
+                is compiled. In addition, compile for different sizes specified
+                in inductor_compile_sizes, using configurations
+                in inductor_compile_config.
+        - inductor_compile_sizes: sizes to compile for inductor.
+        - inductor_specialize_for_cudagraph_no_more_than: an optional integer
+            to specialize inductor for cudagraph sizes no more than the
+            specified size. It is useful when we want to specialize inductor
+            with a subset of cudagraph sizes.
+        - inductor_compile_config: additional configurations for inductor.
+            - None: use default configurations.
+        - inductor_passes: additional passes for inductor. It is a dictionary
+            from pass name to pass function qualified name. We use function
+            name because the config uses json format. If we pass the config
+            from Python, functions can also be passed directly via Python object
+            constructor, e.g. `CompilationConfig(inductor_passes={"a": func})`
+        - custom inductor passes:
+            - dump_graph_stages: list of stages for which we want to dump the graph.
+                Each pass defines its own stages (before, after, maybe in-between).
+            - dump_graph_dir: directory to dump the graph. Default is .
+            - enable_fusion: whether to enable the custom fusion pass.
+                TODO better pass enabling system.
+    
+    Why we have different sizes for cudagraph and inductor:
+    - cudagraph: a cudagraph captured for a specific size can only be used
+        for the same size. We need to capture all the sizes we want to use.
+    - inductor: a graph compiled by inductor for a general shape can be used
+        for different sizes. Inductor can also compile for specific sizes,
+        where it can have more information to optimize the graph with fully
+        static shapes. However, we find the general shape compilation is
+        sufficient for most cases. It might be beneficial to compile for
+        certain small batchsizes, where inductor is good at optimizing.
+    """ # noqa
+    level: int = 0
+    custom_ops: List[str] = Field(default_factory=list)
+
+    use_inductor: bool = True
+    inductor_specialize_for_cudagraph_no_more_than: Optional[int] = None
+    inductor_compile_sizes: Optional[List[int]] = Field(default_factory=dict)
+    inductor_compile_config: Dict = Field(default_factory=dict)
+    inductor_passes: Dict[str, str] = Field(default_factory=dict)
+
+    use_cudagraph: bool = False
+    non_cudagraph_ops: List[str] = Field(default_factory=list)
+    cudagraph_num_of_warmups: int = 0
+    cudagraph_capture_sizes: Optional[List[int]] = None
+    cudagraph_copy_inputs: bool = False
+
+    dump_graph_stages: List[str] = Field(default_factory=list)
+    dump_graph_dir: Path = Field(default=Path("."))
+    enable_fusion: bool = True
+
+    # not configurable, computed after init
+    compile_sizes: List[int] = PrivateAttr
+    capture_sizes: List[int] = PrivateAttr
+
+    def model_post_init(self, __context: Any) -> None:
+        self.level = envs.VLLM_TORCH_COMPILE_LEVEL
+
+        count_none = self.custom_ops.count("none")
+        count_all = self.custom_ops.count("all")
+        assert count_none + count_all <= 1, "Can only specify 'none' or 'all'"
+
+        for k, v in self.inductor_passes.items():
+            if not isinstance(v, str):
+                assert callable(v), (
+                    f"pass {k} should be a function or a qualified name")
+                self.inductor_compile_config[k] = v
+                continue
+
+            # resolve function from qualified name
+            names = v.split(".")
+            module = ".".join(names[:-1])
+            func_name = names[-1]
+            func = __import__(module).__dict__[func_name]
+            self.inductor_compile_config[k] = func
+
+    def init_during_runtime(self):
+        """To complete the initialization of config,
+        we need to know the compile context, which is only available
+        during the first run of the model.
+        """
+        from vllm.compilation.compile_context import get_compile_context
+        context = get_compile_context()
+        context = copy.deepcopy(context) if context is not None else []
+        sizes_to_specialize: List[int] = context
+        if self.cudagraph_capture_sizes is None:
+            self.capture_sizes = sizes_to_specialize
+        else:
+            self.capture_sizes = self.cudagraph_capture_sizes
+            logger.info(("cudagraph sizes specified by model runner"
+                         " %s is overridden by config %s"),
+                        sizes_to_specialize, self.cudagraph_capture_sizes)
+        if self.inductor_specialize_for_cudagraph_no_more_than is not None:
+            assert self.inductor_compile_sizes is None, (
+                "inductor_compile_sizes should be None when "
+                "inductor_specialize_for_cudagraph_no_more_than is not None")
+            self.compile_sizes = [
+                x for x in self.capture_sizes
+                if x <= self.inductor_specialize_for_cudagraph_no_more_than
+            ]
+        else:
+            assert self.inductor_compile_sizes is not None, (
+                "inductor_compile_sizes should not be None when "
+                "inductor_specialize_for_cudagraph_no_more_than is None")
+            self.compile_sizes = self.inductor_compile_sizes
+
+    @staticmethod
+    def select_and_init_config() -> "CompilationConfig":
+        """The order of selecting config is:
+        1. Use the config specified in environment variable.
+        2. Use the config specified in plugins.
+        3. Use the default config.
+        """
+        config_path = envs.VLLM_TORCH_COMPILE_CONFIG
+        if config_path is not None:
+            with open(config_path) as json_file:
+                config = CompilationConfig.model_validate_json(
+                    json_file.read())
+        else:
+            from vllm.plugins import get_compilation_config
+            predefined_config = get_compilation_config()
+            config = predefined_config if predefined_config is not None else (
+                CompilationConfig())
+
+        return config
+
+
 @dataclass
 class VllmConfig:
     """Dataclass which contains all vllm-related configuration. This
@@ -2073,6 +2254,8 @@ class VllmConfig:
     observability_config: Optional[ObservabilityConfig] = None
     prompt_adapter_config: Optional[PromptAdapterConfig] = None
     quant_config: Optional[QuantizationConfig] = None
+    compilation_config: CompilationConfig = field(default=None,
+                                                  init=True)  # type: ignore
 
     @staticmethod
     def _get_quantization_config(
@@ -2133,6 +2316,12 @@ class VllmConfig:
             self.quant_config = VllmConfig._get_quantization_config(
                 self.model_config, self.load_config)
 
+        if self.compilation_config is None:
+            self.compilation_config = CompilationConfig.select_and_init_config(
+            )
+
+        current_platform.check_and_update_config(self)
+
     def __str__(self):
         return ("model=%r, speculative_config=%r, tokenizer=%r, "
         "skip_tokenizer_init=%s, tokenizer_mode=%s, revision=%s, "

vllm/envs.py

@@ -69,7 +69,6 @@ if TYPE_CHECKING:
     VLLM_SKIP_P2P_CHECK: bool = False
     VLLM_TORCH_COMPILE_LEVEL: int = 0
     VLLM_TORCH_COMPILE_CONFIG: Optional[str] = None
-    VLLM_CUSTOM_OPS: List[str] = []
     VLLM_DISABLED_KERNELS: List[str] = []
     VLLM_USE_V1: bool = False
     VLLM_ENABLE_V1_MULTIPROCESSING: bool = False
@@ -217,18 +216,6 @@ environment_variables: Dict[str, Callable[[], Any]] = {
     "VLLM_TORCH_COMPILE_CONFIG":
     lambda: os.environ.get("VLLM_TORCH_COMPILE_CONFIG", None),
 
-    # Fine-grained control over which custom ops to enable/disable.
-    # Use 'all' to enable all, 'none' to disable all.
-    # Also specify a list of custom op names to enable (prefixed with a '+'),
-    # or disable (prefixed with a '-').
-    # Examples:
-    # - 'all,-op1' to enable all except op1
-    # - 'none,+op1,+op2' to enable only op1 and op2
-    # By default, all custom ops are enabled when running without Inductor
-    # and disabled when running with Inductor (compile_level >= Inductor).
-    "VLLM_CUSTOM_OPS":
-    lambda: os.environ.get("VLLM_CUSTOM_OPS", "").replace(" ", "").split(","),
-
     # local rank of the process in the distributed setting, used to determine
     # the GPU device id
     "LOCAL_RANK":

vllm/model_executor/custom_op.py

-from functools import lru_cache
 from typing import Dict, Type
 
 import torch.nn as nn
 
-import vllm.envs as envs
-from vllm.compilation.levels import CompilationLevel
 from vllm.logger import init_logger
 from vllm.platforms import current_platform
+from vllm.plugins import get_current_vllm_config
 from vllm.utils import print_warning_once
 
 logger = init_logger(__name__)
@@ -87,6 +85,8 @@ class CustomOp(nn.Module):
     @classmethod
     def enabled(cls) -> bool:
         # if no name, then it was not registered
+        compilation_config = get_current_vllm_config().compilation_config
+        custom_ops = compilation_config.custom_ops
         if not hasattr(cls, "name"):
             print_warning_once(
                 f"Custom op {cls.__name__} was not registered, "
@@ -94,22 +94,25 @@ class CustomOp(nn.Module):
                 f"It will be enabled/disabled based on the global settings.")
             return CustomOp.default_on()
 
-        enabled = f"+{cls.name}" in envs.VLLM_CUSTOM_OPS
-        disabled = f"-{cls.name}" in envs.VLLM_CUSTOM_OPS
+        enabled = f"+{cls.name}" in custom_ops
+        disabled = f"-{cls.name}" in custom_ops
         assert not (enabled
                     and disabled), f"Cannot enable and disable {cls.name}"
 
         return (CustomOp.default_on() or enabled) and not disabled
 
-    # On by default if VLLM_TORCH_COMPILE_LEVEL < CompilationLevel.PIECEWISE
-    # Specifying 'all' or 'none' in VLLM_CUSTOM_OPS takes precedence.
     @staticmethod
-    @lru_cache
     def default_on() -> bool:
-        count_none = envs.VLLM_CUSTOM_OPS.count("none")
-        count_all = envs.VLLM_CUSTOM_OPS.count("all")
-        assert count_none + count_all <= 1, "Can only specify 'none' or 'all'"
-        return envs.VLLM_TORCH_COMPILE_LEVEL < CompilationLevel.PIECEWISE and \
+        """
+        On by default if level < CompilationLevel.PIECEWISE
+        Specifying 'all' or 'none' in custom_op takes precedence.
+        """
+        from vllm.config import CompilationLevel
+        compilation_config = get_current_vllm_config().compilation_config
+        custom_ops = compilation_config.custom_ops
+        count_none = custom_ops.count("none")
+        count_all = custom_ops.count("all")
+        return compilation_config.level < CompilationLevel.PIECEWISE and \
             not count_none > 0 or count_all > 0
 
     # Dictionary of all custom ops (classes, indexed by registered name).